Proc. Natl. Acad. Sci. USA Vol. 73, No. 6, pp. 1897-1901, June 1976 Biochemistry Solenoidal model for superstructure in chromatin (electron microscopy/packing ratio/supercoil) J. T. FINCH AND A. KLUG MRC Laboratory of Molecular Biology, Cambridge CB2 2QH, England Communicated by F. H. C. Crick, March 18,1976

ABSTRACT Chromatin prepared by brief digestion of nu- matin and that nucleofilaments are normally organized in a clei with micrococcal nuclease, and extracted in 0.2 mM EDTA, higher level of folding. appears in the electron microscope as filaments of about 100 In this paper we present evidence obtained with the electron X diameter which coil loosely. In 0.2 mM Mg++ these "nucleo- filaments" condense into a supercoil or solenoidal structure of microscope for a state of chromatin in which nucleofilaments pitch about 110 A corresponding to the diameter of a nucleofi- are folded into supercoils or solenoidal structures of pitch about lament. It is proposed that the x-ray reflections at orders of 110 110 A. In our experiments we have used chromatin prepared A observed in chromatin originate in the spacing between turns by a brief micrococcal nuclease digestion of nuclei, which is of the solenoid rather than that between nucleosomes along the "native" in the sense that further nuclease digestion gives the nucleofflament. Ihe solenoidal structure appears to need histone same discrete pattern as does direct digestion of nuclei (15). HI for its stabilization. Under certain conditions, isolated nu- cleosomes can also aggregate into a similar structure. The sol- Chromatin prepared by methods that involve shearing does not enoidal structure can be correlated with the "thread" of diam- give such clear digestion patterns. In the presence of chelating eter about 300 A observed by other workers in nuclei. agents, this native chromatin appears as fairly uniform fila- ments of 100 A diameter, which are curly or loosely coiled. In The results of biochemical and other studies by various workers the presence of Mg++ ions, tightly coiled forms of the nucleo- (1-4) led Kornberg to propose a model for the basic structure filament are dominant, leading to a supercoil or solenoidal of chromatin (5). He suggested that it consists of a flexible chain structure in which the pitch corresponds to the diameter of a of repeating structural units of about 100 A diameter ("beads"), nucleofilament. each containing a stretch of DNA 200 base pairs long condensed around a protein core made out of eight histone molecules (two MATERIALS AND METHODS each of the four main types). These elementary units have been named "nucleosomes" (6). The physicochemical 'parameters Chromatin was prepared from rat liver nuclei after brief mi- for the model require that the nucleosomes are in close contact crococcal nuclease digestion, as described previously (15), to along the chain, and there is now good evidence from electron give an average length of about 40 nucleosomes, with a range microscopy that this is the case (7, 8). We call this close-packed roughly from 10 to 100. chain a "nucleofilament." A recent study by low angle x-ray Samples from dilute solutions of chromatin in appropriate scattering of dilute solutions of chromatin, prepared without media (e.g., 0.2 mM EDTA or 0.2 mM MgCl2) were applied to shearing, has given a mass per unit length consistent with nu- carbon-coated grids and washed with the same medium. At the cleosomes densely packed along the length of a nucleofilament typical chromatin concentrations used, a solution containing (9). 0.1 mM Mg++ is just sufficient to neutralize the phosphate Linear arrangements of beads in a partly extended form groups. For negative staining, the grids were then washed with ("beads-on-a-string") were first reported by Olins and Olins a few drops of 1% uranyl acetate, the excess being withdrawn (10) and Woodcock (11) in chromatin spilling out of nuclei lysed onto a filter paper. For shadowing of freeze-etched specimens, on the specimen grid, and made visible by negative or positive the specimens were fixed by floating the support grid on a drop staining. This extended form of chromatin probably results of 2% formaldehyde in the same solution for a few minutes and from the shearing forces present during the nuclear explosion. then were wasbed by floating on water. The specimens were A similar extended appearance is seen in the electron micro- then rapidly frozen in liquid nitrogen, dried on a Balzer's ap- graphs obtained by Oudet et al. from extracted chromatin, after paratus, and shadowed with platinum-carbon at an angle of 350, removal of histone HI (6). However, electron micrographs of as described by Nermut (16). small oligomers of beads prepared from nuclei by brief nuclease HI-depleted chromatin was prepared by treating chromatin digestion showed a preference for the nucleosomes to be in prepared as above with 0.45 M or 0.6 M NaCI, and dialyzing contact (8). against Tris buffer (5 mM Tris-acetate, 20 mM ammonium The nucleofilament can be correlated with the fairly uniform acetate, pH 6.8) after isolation on a sucrose gradient. Under fibers or threads of 100 A diameter seen in the electron mi- these conditions, virtually all the histone Hi was removed with croscope in chromatin specimens prepared by critical point almost no concomitant loss of the other four histones (L. Lutter, drying in the presence of chelating agents (4). However, spec- unpublished work). imens prepared in the absence of chelating agents show thicker fibers of about 250 A diameter (4, 12), suggesting that chro- RESULTS matin is normally present in a higher level of organization that breaks down when divalent ions are removed. The appearance Electron microscope observations of the thinner fibers has been attributed to the unwinding of Electron micrographs of chromatin in 0.2 mM EDTA or in 0.1 tertiary coiling (13, 14) or to the separation of two associated mM Mg++ are shown in Figs. 1 and 2. Chromatin appears as 100 A fibers (4, 14). Another line of reasoning, based on x-ray filaments about 100 A in diameter, which often tend to coil up studies and described below, also suggests that Kornberg's linear loosely. There is no visible repeat along the length of a filament. chain model only represents the first level of folding in chro- The filamentous character agrees with Sperling and Tardieu's 1897 Downloaded by guest on September 25, 2021 1898 Biochemistry: Finch and Klug Proc. Natl. Acad. Sci. USA 73 (1976)

.s

FIG. 2. Chromatin in 0.2 mM EDTA freeze etched and shad owed. The roughly parallel orientation of filaments probably arises during adsorption or washing of the specimen on the grid. X140,000.

at least 300 'A thick, consistent with a helical character, rather than a two-dimensional zig-zag type of folding. FIG. 1. Electron micrographs of negatively stained chromatin The direction of the transverse striations in shadowed spec- prepared in (a) 0.2 mM EDTA, or (b) 0.1 mM Mg++. Arrows indicate imens does not consistently indicate a right- or left-handed helix, regions where loose coiling is evident. X140,000. although only one surface is being viewed. Since the slope of the helix is small, its left- or right-hand direction on one side of the fiber would be very susceptible to shear, and such shearing results on low angle x-ray scattering from dilute solutions of this is quite often present, as judged by the changes in obliquity of material in the same medium, which are consistent with the the striations seen in both types of specimens. Indeed, the scattering from a uniform cylinder of about 100 A in diameter variations in diameter and occasional pulled out fibers show that (9). the forces responsible for setting the relative phase of successive In a few instances (Fig. lb) the coiling of the nucleofilaments turns of the supercoil are not that strong or specific, at least for is so pronounced as to lead to a more condensed state, but the this type of preparation. Similarly, the helical parameters other majority of filaments are extended. In 0.2 mM MgCl2, however, than pitch are difficult to determine. The lack of visible detail almost all the material has this condensed character. The con- along the turns 'of the solenoids and the variation in diameter centration of Mg++ is not critical above 0.2 mM; similar results make it difficult to judge the number of nucleosomes per turn, are obtained up to 1 mM Mg++. In images of both negatively but the mean appears to be about 6 or 7. The variation in di- stained and shadowed specimens (Figs. 3 and 4) thick fibers or ameter may result from the relatively short length of the pieces elongated particles can be seen, of a diameter that is somewhat of chromatin used (about 40 nucleosomes), which correspond variable, even along the same fiber, but is usually in the range to only six turns of helix on the average. 300-500 A. In the negatively stained images striations can fairly This tendency for an association involving the "sides" of a often be seen running across the fibers, with a spacing of about nucleofilament is also shown, under certain circumstances, by 120-150 A. We regard the smaller values as the more signifi- individual nucleosomes. Aggregates with an appearance similar cant, since the longer spacings occur when the fibers are pulled to that of the Mg++-induced fibers have been obtained from out. The striations become increasingly difficult to detect as solutions of isolated nucleosomes (prepared as in ref. 17), which their spacing decreases, indicating the possibility of still tighter had been taken off a sucrose gradient and dialyzed against packing. deionized water (Fig. 5). This suggests that this mode of Similar striations can be seen in the shadowed specimens aggregation is a natural one between nucleosomes and not de- when the shadow direction is favorable, but there is little in- pendent on the continuity of the DNA along a nucleofilament. dication of further substructure in the transverse direction across An electron micrograph of Hi-depleted chromatin is shown the fiber width. These observations suggest that the fibers are in Fig. 6a. The particles no longer have the appearance of a formed by the winding up of the nucleofilaments into helices. continuous filament but rather of a chain of discrete beads, The lengths of the shadows in Fig. 4 indicate that the fibers are looking very like the micrographs published by Oudet et al. (6) Downloaded by guest on September 25, 2021 Biochemistry: Finch and Klug Proc. Nati. Acad. Sci. USA 73 (1976) 1899

FIG. 5. Aggregates of nticleosomes in water, negatively stained. The nucleosomes were prepared and separated as described in ref. 17, and dialyzed for 24 hr against deionized water. The aggregates dissociate into individual nucleosomes on addition of 25 mM am- monium acetate, and are not found in the presence of 0.2 mM EDTA (8). X140,000.

FIG. 3. Chromatin in 0.5 mM Mg++, negatively stained. Arrows It would seem natural to associate this 110 A x-ray spacing indicate transverse striations across elongated particles. X 140,000. with the diameter of a nucleosome (5), and a discussion of its origin in terms of the packing of 110 A spheres has recently been given (22). However, the 110 A series of x-ray reflections cannot for HI-depleted chromatin. Addition of Mg++ (Fig. 6b) does arise from the interval between nucleosomes along the nu- not lead to condensed forms of the type described above: dis- cleofilament direction, since the x-ray scattering curve in the crete beads are still evident, although the pieces of chromatin region of spacings of 400-30 A from nucleofilaments in dilute form more localized patches than without Mg++. Similar results solution in 0.2 mM EDTA (9) fits the theoretical curve for a were obtained with chromatin depleted of HI by use of tRNA continuous density rod, with no signs of peaks at 110 A or 55 A. (18). It is not clear whether removal of HI produces a more This is consistent with the lack of contrast seen along the nu- extended form in solution, which will not condense into a su- cleofilaments in the electron micrographs described above. percoil, or whether its absence merely destabilizes the material Instead, we propose that the 110 A spacing originates as a on the grid of the electron microscope. In a very recent sedi- spacing between nucleofilaments organized in a higher level mentation analysis of pieces of chromatin up to 10 nucleosomes of folding. The simplest form of such a folding is a supercoil or long, M. Noll and R. D. Kornberg (unpublished work) have solenoid with a pitch of 110iA, and the electron micrographs found that removal of HI causes a marked reduction in sedi- shown above are evidence for a state of chromatin in which the mentation constant, consistent with the opening out of a more nucleofilament is folded into such solenoidal structures. Our compact structure. It would thus seem likely that Hi is needed interpretation of the x-ray pattern is therefore that the 110 A for the formation or stabilization of the solenoidal structure. reflection and its higher orders arise from the spacing between Relation to x-ray diffraction studies the turns of a solenoidal type of structure. Evidence for this interpretation has been provided by our X-ray diffraction patterns of chromatin generally show a series colleague, L. Sperling, who has taken x-ray photographs of of diffraction maxima or bands, nominally at 110, 55, 37, 27, and 22 A, as originally described by Luzzati and Nicolaieff (19, 20) and by Wilkins et al. (3). The pattern becomes sharper and a *'., .b emerges from the background as the concentration is raised, becoming distinct at concentrations higher than about 35% (wt/wt). X-ray studies on fibers of chromatin have shown these maxima to be oriented in the fiber direction (21).

FIG. 4. Chromatin in 0.2 mM Mg++, freeze-etched and shadowed. The slope of the striations is not consistent on all particles, probably FIG. 6. (a) Negatively stained chromatin in Tris buffer, after because of distortion of the rather short particles, but there seems to depletion of histone H1 by salt treatment. (b) The same preparation be a preference for a left-handed helix. X14Q,000. in 1 mM Mg++. X75,000. Downloaded by guest on September 25, 2021 1900 Biochemistry: Finch and Klug Proc. Natl. Acad. Sci. USA 73 (1976) number are seen with a pitch angle large enough to correspond to a two-start helix, but as mentioned earlier, this appearance might arise by shearing of a simple helix. Is there any evidence that the solenoidal structure actually exists in chromosomes, apart from the observations, referred to above, that "thick" (i.e., 250 A diameter) fibers can be ex- tracted from nuclei? The best evidence comes from the work of Davies and his colleagues, who have shown the presence of thread-like units in condensed interphase chromosomes or chromatin bodies. These tend to line up in layers at the surface of the nucleus, and the monolayers from a wide range of species have an average width of about 300 A (25, 26). it ]- 110A Previously was supposed that threads roughly 170 A in diameter were em- bedded in a matrix (27), but the latest interpretation of the staining characteristics (25) is that an inner rich DNA-core of this diameter is surrounded by an outer shell of external di- FIG. 7. Schematic diagram showing the folding of a nucleofila- ameter 280 A in which the protein-to-DNA ratio is higher. More ment into a solenoid. The thin line shown wound as a helix along the recently it is suggested that the thread-like unit arises from some nucleofilament is intended to represent the folding of the DNA type of helical arrangement of the "beads-on-a-string" (28). double helix on the outside of a protein core (5, 24, 8); it is highly It schematic, since the path or fold is not known. would seem natural to correlate this "superunit" thread (25) with our solenoidal model, and to take a value of about 300 A as an estimate for the outer diameter. Allowing for possibility centrifuge pellets of material that showed solenoidal structures of shrinkage in the specimens as prepared for the electron mi- in the electron microscope. Although only at concentrations of croscope, the solenoid in its native state would then have about 25% (wt/wt) or less, this material gave the characteristic type six nucleosomes per turn of the helical coil. A bare nucleofila- of x-ray pattern consisting of bands at orders of I10 A; it showed ment consisting only of elementary nucleosomes wound into more order than hitherto had been obtained, even at concen- such a coil would leave a central hole of diameter about 100 A. trations of 40% or over. This x-ray work will be described Thus, the solenoidal structure observed does not represent the elsewhere (Sperling and Klug, in preparation), but briefly the closest packing of nucleosomes. The closest packing of spheres character of the pattern is consistent with its arising from the wound into a helix leaving no hole down the center would have turns of solenoids of low pitch angle. Indeed, our explanation about 2.7 units per turn (29). Presumably in chromosomes, the for the x-ray pattern would be of the type originally proposed central hole in a solenoid might well be occupied by histone Hi by Pardon and Wilkins, when they suggested their "supercoil" and some of the nonhistone proteins or other control elements. model for the DNA in chromatin (23). However, the structure It has been suggested to us by Dr. F. H. C. Crick that this central we propose is at a higher level of folding, the path of the su- hole might also accommodate some of the "naked" DNA shown percoil now being followed by the nucleofilament rather than to exist as a small proportion of the total DNA in chromosomes the double helix of DNA. after removal of histone HI (30, 6). This would be in accord We thus interpret earlier x-ray studies carried out without with Davies' suggestion of a DNA-rich core to explain the control of the Mg++ concentration as follows. The regular pe- preference for uranyl-lead stain to pick out the centers of the riodicity in the x-ray patterns and its variation with increasing threads, although this preference might also be accounted for concentrations of chromatin results from the formation of sol- by occlusion of the stain in a cylindrical cavity whose walls enoidal structures on the withdrawal of water. At low concen- would necessarily be coated by the DNA wrapping around the trations the nucleofilament tends to be extended, and as the nucleofilament. concentration is raised, more or less stable turns of the solenoid The packing or compaction ratio for the DNA along the are formed. On this picture there is no necessity for a constant length of a densely packed nucleofilament is about 7:1 (5), and spacing between turns, and one might expect a gradual decrease when the nucleofilament itself is wound into a solenoid of the in this spacing and also an increase in order, as the solenoid type suggested, this packing ratio will increase to about 40:1 in tightens. In fact, the x-ray spacings decrease by about 10% as the direction of the solenoid axis. Higher packing ratios can of the chromatin concentration is increased from 30 to 50% course by achieved by coiling the solenoidal form of the nu- (wt/wt) (Kornberg, Klug, and Sperling, unpublished work). cleofilament into yet another order of helix, or by arranging a number of solenoids to wind around each other. It has been DISCUSSION pointed out (31) that the construction of a hierarchy of helices The solenoidal form of chromatin is represented schematically can be achieved in a neat way by allowing the DNA double in Fig. 7. For definiteness the nucleofilament is drawn as folded helix to kink locally rather than bend or distort over stretches, into a coil containing about six nucleosomes per turn, but our just as postulated for the folding of the DNA on the nucleosome observations in the electron microscope do not establish this itself. number. We have observed a fairly wide range of widths of the We our solenoidal particles, some of which variation is undoubtedly due thank colleagues Drs. R. D. Kornberg, M. Noll, L. Lutter, and especially Ms L. Sperling for supplying us with chromatin and for to flattening, but would seem to represent a range of coils advice. We thank Dr. F. H. C. Crick for many helpful discussions. containing from 4 to 10 units per turn, with a mode at about 6 or 7. under the our Presumably, conditions of experiments there 1. Hewish, D. R. & Burgoyne, L. A. (1973) Biochem. Biophys. Res. is a family of such solenoids all of nearly the same pitch (brought Commun. 52,504-510. about by contacts involving the "sides" of nucleosomes) but with 2. Kornberg, R. D. & Thomas, J. 0. (1974) Science 184, 865- a varying number of units per turn. In most cases, the pitch 868. angle is small, suggesting a single-start helix. However, a 3. Wilkins, M. H. F., Zubay, G. & Wilson, H. R. (1959) J. Mol. Biol. Downloaded by guest on September 25, 2021 Biochemistry: Finch and KMug Proc. Nat!. Acad. Sci. USA 73 (1976) 1901

1,179-185. & Favard, P. (Societe Francaise de Microscopie Electronique, 4. Ris, H. & Kubai, D. F. (1970) Annu. Rev. Genet. 4,263-294. Paris), Chap. 12. 5. Kornberg, R. D. (1974) Science 184,868-871. 17. Noll, M. (1974) Nature 251,249-251. 6. Oudet, P., Gross-Bellard, M. & Chambon, P. (1975) Cell 4, 18. Ilyin, Y. V., Varshavsky, A. V., Mickelsaar, U. N. & Georgiev, G. 281-300. P. (1971) Eur. J. Biochem. 22,235. 7. Griffith, J. D. (1975) Science 187,1202-1203. 19. Luzzati, V. & Nicolaieff, A. (1959) J. Mol. Biol. 1, 127-133. 8. Finch, J. T., Noll, M. & Kornberg, R. D. (1975) Proc. Natl. Acad. 20. Luzzati, V. & Nicolaieff, A. (1963) J. Mol. Biol. 7, 142-163. Sci. USA 72,3320-322. 21. Pardon, J. F., Richards, B. M. & Cotter, R. I. (1974) Cold Spring 9. Sperling, L. & Tardieu, A. (1976) FEBS Lett. 64, 89-91. Harbor Symp. Quant. Biol. 38, 75-81. 10. Olins, A. L. & Olins, D. E. (1974) Science 183,330-332. 22. Carlson, R. D. & Olins, D. E. (1976) Nucleic Acid Res. 3,89-100. 11. Woodcock, C. L. F. (1973) J. Cell Biol. 59, 368a. 23. Pardon, J. F. & Wilkins, M. H. F. (1972) J. Mol. Biol. 68,115-124. 12. Pooley, A. S., Pardon, J. F. & Richards, B. M. (1974) J. Mol. Biol. 24. Noll, M. (1974) Nucleic Acid Res. 1, 1573-1578. 85,533-549. 25. Davies, H. G., Murray, A. B. & Walmsley, M. E. (1974) J. Cell 13. Dupraw, E. J. (1970) in DNA and Chromosomes (Holt, Rinehart Sci. 16, 261-299. & Winston, New York), Chap. 10. 26. Davies, H. G. & Haynes, M. E. (1975) J. Cell Sci. 17, 263-285. 14. Ris, H. (1974) in The Structure and Function of Chromosomes, 27. Davies, H. G. (1968) J. Cell Sci. 3, 129-150. CIBA Foundation Symposium 28 (Associated Scientific Pub- 28. Davies, H. G. & Haynes, M. E. (1976) J. Cell Sci., in press. lishers, Amsterdam), pp. 7-23. 29. Boerdijk, A. H. (1952) Philips Res. Rep. 7,303-313. 15. Noll, M., Thomas, J. 0. & Kornberg, R. D. (1975) Science 187, 30. Varshavsky, A. J., Ilyin, Y. V. & Georgiev, G. P. (1974) Nature 1203-1206. 250,602. 16. Nermut, M. V. (1973) in Freeze Etching, eds. Benedetti, E. L. 31. Crick, F. H. C. & KMug, A. (1975) Nature 255,530-533. Downloaded by guest on September 25, 2021